Statistical Sciences

Primary Expertise and Research Areas

• Analysis of Measurement Systems
• Bayesian Methods
• Design and Analysis of Experiments
• Reliability
• Sampling and Data Planning
• Statistical Computation
• Statistical Graphics and Visualization
• Uncertainty Quantification of Computer Models

Current and Emerging Strategic Thrusts

• Trustworthy and robust AI, including methods for explainability, interpretability, testing, and evaluation
• Surrogate models for scientific computing and AI for advanced property inference, design exploration/optimization, and control on engineered systems
• Statistical and physics-informed machine learning methods to improve computational science algorithms and advance prediction
• Advances in statistical modeling methods, including uncertainty quantification, verification, and validation, to integrate theory, statistical models, and experimental data for multi-systems or multi-physics applications and multi-physics codes
• Statistical methods and techniques to support design and production qualification, testing considerations for production and fielding, and quality improvements across manufacturing and production processes
• Data collection and modeling methods to support the development of new approaches to alternative or modified energy systems and the transition from fossil fuels

Scientific Software

The Statistical Sciences Group develops cutting-edge statistical methods, models, and the algorithms and computer codes to support them. This is a collection of projects that the Statistical Sciences Group has released publicly.

• DeBoinR—Visualization and outlier detection for probability density functions
• FastGP—Efficiently using Gaussian Processes with Rcpp and RcppEigen. Contains Rcpp and RcppEigen implementations of matrix operations useful for Gaussian process models.
• GPM/SA—Gaussian Process Models for Simulation Analysis (GPM/SA) is a Matlab toolset for simulation analysis and uncertainty quantification
• Multiverse—Python package for Bayesian neural network inference, including MCMC and stochastic variational inference with Laplace approximations in development
• Prism—The Programming Repository for In Situ Modeling (PRISM) is a set of tools for fitting statistics and machine learning models to simulation data inside the simulations as they are running 
• Sepia—Simulation-Enabled Prediction, Inference, and Analysis: physics-informed statistical learning. Implements Bayesian emulation and calibration with the ability to handle multivariate outputs
• SplitML—Signal Processing Library for Interference rejecTion by Machine Learning (SplitML) is a Python package for complex-valued signal denoising using statistical and neural network methods